Opinion

Felix Salmon

Sovereign default datapoints of the day

By Felix Salmon
July 6, 2010

CMA Datavision has released its sovereign risk report for the second quarter of 2010. It’s based on CDS prices, and it makes for fascinating reading. It was a bad quarter for sovereign credit: 93% of sovereigns widened, and they widened a lot — by 30%, on average. Spreads in France, Portugal, and Spain all more than doubled, while Greece soared from 346bp to 1,003bp, at one point becoming the riskiest sovereign in the world, trading slightly wider than Venezuela. Only one country saw its CDS spreads tighten in significantly over the second quarter: Iceland tightened in by 17% to 330bp, making it riskier than Ireland but safer than Portugal.

CMA turns all of its CDS data into a 5-year cumulative probability of default: both Venezuela and Greece are more likely than not to default at some point in the next five years, according to these numbers, while Argentina comes very close. Portugal, Latvia, Iceland, Ireland, Spain and Croatia are all in the 20%-25% range, along with Lebanon, which has long had a debt which would be unsustainable were it not for a rich, generous and patriotic diaspora.

One interesting thing is that CDS spreads don’t correspond directly with default probability: Croatia, for instance, has a 20.5% chance of default with CDS spreads at 322bp, while Spain has a higher default probability — 20.7% — while having significantly tighter CDS spread of 265bp.

The reason is that CMA is using at least two different recovery ratios. If Greece or Spain or Italy defaults (or if the US does, for that matter), then CMA assumes that creditors will get back 40% of their money. But the recovery value in countries like Croatia and El Salvador is assumed to be significantly lower, at 25%.

Both figures are low, but I don’t think that this discrepancy makes sense. It seems to me that recovery values should be inversely correlated with debt-to-GDP ratios: if you have relatively little debt, then a modest restructuring can get you back on a sustainable footing, while if you have a lot of debt — like Greece, for instance — then the haircut you’ll need to impose on your creditors is much greater. I see no good reason at all for assuming that recovery values in Croatia will be higher lower than in Greece, especially after accounting for the fact that Greece is likely to have a large number of preferred creditors in Europe who will insist on being paid back in full before private-sector creditors get anything.

I do think that in future reports CMA should make its recovery-value assumptions explicit. The CDS data is important and interesting, but the default probabilities are less so, based as they are on recovery rates which seem dubious indeed.

Comments
6 comments so far | RSS Comments RSS

most sovereign cds (not european though) also trade in a different currency than the country they are insuring, so there is an FX rate assumption built in there as well. this can be quite significant — you are asking what the relative FX rate move would be given a sovereign default. good luck untangling this.

Posted by q_is_too_short | Report as abusive
 

In the fifth paragraph it read “I see no good reason at all for assuming that recovery values in Croatia will be higher than in Greece”. I believe that the CMA assumption is the other way and that you are saying you see no reason to assume the recovery value of Greek debt will be higher than that of Croatia.

Posted by JohnOmeara | Report as abusive
 

Sorry John, well spotted. Fixed.

Posted by FelixSalmon | Report as abusive
 

Part of the widening may reflect an awareness that CDSs are unlikely to be bailed out by sovereigns anymore in the future, and that the great CDS bailout of 2008 is unlikely to be repeated now that the public knows about these instruments.

A strong case is to be made that until recently CDS coverage was *underpriced* because of an implicit government backstop.

Posted by DanHess | Report as abusive
 

1. Iceland imho is already defaulting at the moment, according to international treaties they have to reimburse the UK and Holland (may be the status of part of that amount is somewhat dodgy, but now they are paying nothing at all).
2. Recovery rate standardization is imho totally wrong.
Spain with 50-55% debt giving a haircut which would leave only 20-22% just as an example. It would become the best kid in the block by far. Spain is one of the most important countries in this respect.
3. EU countries have the advantages of a safety-net and a remaining relatively strong own currency (providing they stay in the EU and EURO-zone), post haicut debt is in a ‘normal’ currency.
4. CDS is probably not such a good measure because of the reason mentioned by DanHess, they are relatively expensive. Basically you need collateral and not margin.
5. Volumes of eg Greek debt and CDS are marginal and it doesnot look like the market has found an equilibrium for the new situation yet, with some sellers and very few normal buyers.
6. Influence of ECB programm however would suggest an even higher chance of default for Greece.

Posted by Rikh | Report as abusive
 

OK, suppose God comes down from heaven and informs you that your recovery rate assumption is correct and that you have correctly modeled the joint stochastic evolution of hazard rates and FX. Now you can confidently imply a default probability, but what is the point of the exercise?

What you have is a risk-neutral probability, i.e. the probability that gives you the right price in your chosen units. If your goal is to make inferences about “real” probabilities, this is every bit as circular as it sounds, because it embeds the price of risk. One expects (hopes?) this price is positive, in which case the implied probability must be higher than the “real” one. But if DanHess is right, then the price of risk might be negative, giving you probabilities that are too low – maybe even negative themselves. A negative probability would be a clear indication that some of your assumptions are wrong, but in general you can’t decide whether your probability is too high or too low without knowing the “real” probability in the first place.

Posted by Greycap | Report as abusive
 

Post Your Comment

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see http://blogs.reuters.com/fulldisclosure/2010/09/27/toward-a-more-thoughtful-conversation-on-stories/
  •